lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20140702164629.GI5541@saruman.home>
Date:	Wed, 2 Jul 2014 11:46:30 -0500
From:	Felipe Balbi <balbi@...com>
To:	sundeep subbaraya <sundeep.lkml@...il.com>
CC:	"balbi@...com" <balbi@...com>,
	Alan Stern <stern@...land.harvard.edu>,
	Subbaraya Sundeep Bhatta <subbaraya.sundeep.bhatta@...inx.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Michal Simek <michals@...inx.com>,
	"linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
	Subbaraya Sundeep Bhatta <sbhatta@...inx.com>
Subject: Re: [PATCH v2 2/2] usb: gadget: Add xilinx axi usb2 device support

Hi,

On Sun, May 25, 2014 at 11:10:30PM +0530, sundeep subbaraya wrote:
> Hi Felipe,
> 
> Please take a look at below about how this IP works:
> 
> IN:
>       req.buf ---> DMA (transfers from ddr to IP buffer, raise DMA
> done interrupt and set Buffer ready to transfer data to Host)---->Host
> PC
>   buffer sent interrupt
> 
> OUT:
>      Host PC--->buffer ready interrupt--->DMA (transfer from IP buffer
> to DDR,DMA done interrupt, set Buffer ready to receive next data from
> Host)-->req.buf
> 
> I written logic to call completion in DMA done handler because it
> works for both IN and OUT eps. But I see significant performance
> degradation (by copying a file to mass storage gadget). DMA can handle
> unaligned address too but it has ep max limit (say 512 for bulk).
> Hence it is SW job to split packets and to drive DMA till req.length
> completes. I feel polling for a while is faster than switching between
> interrupt handler back and forth rapidly. Moreover, DMA may or may not
> be present in IP based on user configuration at design time.
> 
> I fixed all other comments expect this one if you do not agree for
> polling then I may have to create threads and do this stuff. What do
> you suggest?

Can you resend your driver so we can review again everything you have
fixed ? I guess you have no other way... polling it is.

cheers

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ